Phenotype linkage-based rule mining for polypharmacological drug discovery

Abstract

Polypharmacology has been the important research area due to the limitations of the conventional pharmacology. The phenotypic evidence-based approaches are promising with advantages of utilizing the accumulated empirical knowledge about efficacy and safety of existing medicines. However, the conventional approaches did not consider the phenotype linkage, which the various relatedness among them. In this study, we propose a novel framework of rule mining based on the phenotype linkage to discover potential therapeutic compound combinations in the context of polypharmacological drug discovery. For the purpose, we construct an integrated database of comprehensive information of medicinal materials. We introduce the efficacy score of medicinal materials that quantifies therapeutic effects on a target disease with considering the phenotype linkage, and design a framework of rule mining from medicinal knowledge bases. The proposed method show the strength in reproducibility and reliability, when compared with the conventional rule mining method. Furthermore, we discover the potential therapeutic compound combinations frequently contained in medicinal herbs that are highly related with target diseases.